THE IMPACT OF CLIMATE CHANGE ON WILDFIRE

WORKSTREAM 4: RESEARCH REPORT

2019

2

Authors Greg Forsyth and David Le Maitre

Date 2019

ToDB reference

Suggested citation Forsyth, G. & Le Maitre, D. 2019. Green Book – The impact of climate change on wildfires. Technical report, Pretoria: CSIR

Disclaimer and acknowledgement: This work was carried out with the aid of a grant from the CSIR Long-term

Thematic Programme, Pretoria, South Africa and the International Development Research Centre, Ottawa,

Canada. The views expressed herein do not necessarily represent those of the IDRC or its Board of Governors.

CSIR/NRE/ECOS/ER/2019/0003/C

3

tents

1 INTRODUCTION ............................................................................................. 7

2 BACKGROUND .............................................................................................. 8

2.1 Fire hazard ............................................................................................................ 8

2.2 Changes in land-cover and fire hazard ................................................................ 10

2.3 Fire vulnerability .................................................................................................. 12

2.4 Fire risk................................................................................................................ 13

2.5 Modelling fire danger ........................................................................................... 14

2.5.1 Fynbos and Renosterveld ............................................................................ 14

2.5.2 Woodland and Grassland ............................................................................. 15

3 METHODOLOGY .......................................................................................... 17

3.1 Defining the area for assessing the hazard .......................................................... 17

3.2 Characterising fuels and fire hazard .................................................................... 18

3.3 Quantifying the fire hazard for each settlement .................................................... 26

3.4 Determining the consequences of a wildfire for each settlement .......................... 26

3.5 Determining the final level of social and economic consequence for each settlement

29

3.6 Using a risk matrix to assign a fire risk rating ....................................................... 30

3.7 Accounting for the effects of climate change on fire danger ................................. 31

3.8 Using records for fire occurrence ......................................................................... 31

4 RESULTS ...................................................................................................... 31

4.1 Fire hazard or likelihood ...................................................................................... 31

TABLE OF CONTENTS

4

4.2 Economic and social consequences .................................................................... 34

4.3 Fire-danger days under current and future climates ............................................. 37

4.4 Cederberg Local Municipality ............................................................................... 40

4.5 Dihlabeng Local Municipality ............................................................................... 42

4.6 Fire occurrence .................................................................................................... 46

5 CONCLUSION .............................................................................................. 48

6 REFERENCES .............................................................................................. 49

TABLE OF FIGURES

Figure 1: Fire ecology types and their distribution in South Africa, Lesotho and Swaziland

(Bond, 1997; Forsyth et al., 2010; Le Maitre et al., 2014b). ................................................... 9

Figure 2: Total burnt area (expressed as proportion of total area within each vegetation type)

in the (a) KZN Sandstone Sourveld, (b) Ngongoni Veld, (c) Eastern Valley Bushveld and (d)

KZN Coastal Belt over a 10-year period (from Buthelezi et al., 2016). ................................. 16

Figure 3: Percentage of the KwaZulu-Natal Sandstone Sourveld, Ngongoni Veld, Eastern

Valley Bushveld and KwaZulu-Natal Coastal Belt that burnt in each month (March to

September) (from Buthelezi et al., 2016). ............................................................................ 17

Figure 4: Spatial variations in the likelihood of wildfires occurring within the exterior 1 km buffer

on WUI for Bethlehem and Clarens in the eastern Free State. ............................................ 34

Figure 5: Land-cover classes present in the internal buffer of the Wildland-Urban Interface of

a single South Africa settlement based on the 2013-14 national land-cover (GTI, 2015). .... 35

Figure 6: Land-cover classes for a single town categorised by levels of economic

consequence present within the internal buffer of the Wildland-Urban Interface of a single

South Africa settlement. ...................................................................................................... 36

Figure 7: Economic risks aggregated to the dominant risk class occurring within the interior

buffer of the Wildland-Urban Interface of a single town in South Africa. .............................. 37

Figure 8: Classes of very high fire-danger days for southern Africa under the current climate

(1961-1990). ....................................................................................................................... 38

Figure 9: Classes of very high fire-danger days for southern Africa under the climate projects

for the near future (2021-2050). .......................................................................................... 39

5

Figure 10: Classes of very high fire-danger days for southern Africa under the climate projects

for the far future (2021-2050). ............................................................................................. 39

Figure 11: The Cederberg Local Municipality showing the location of the important settlements

and the variable topography from the nearly flat West Coast lowlands to the rugged interior,

particularly the Cederberg Mountains east of Clanwilliam. .................................................. 40

Figure 12: The number of very high fire-danger days for each the towns within the Cederberg

Municipality under the current climate (1961 to 1990 .......................................................... 41

Figure 13: The number of very high fire-danger days predicted for the near future (2021 to

2050) for each of the towns within the Cederberg Municipality. ........................................... 41

Figure 14: The number of very high fire-danger days predicted for the far future (2070 to 2099)

for each the towns within the Cederberg Municipality. ......................................................... 42

Figure 15: Dihlabeng Municipality in the eastern Free State showing the main town of

Bethlehem and the other small settlements. The southern boundary of the municipality is the

border with Lesotho. ........................................................................................................... 42

Figure 16: The fire hazard (likelihood) of the different settlements in the Dihlabeng Municipality

in the eastern Free State. .................................................................................................... 43

Figure 17: The economic consequences of wildfires in the settlements in the Dihlabeng

Municipality in the eastern Free State. ................................................................................ 43

Figure 18: The economic risk associated with the occurrence of a wildfire in the settlements

in the Dihlabeng Municipality in the eastern Free State. ...................................................... 44

Figure 19: The number of very high fire-danger days for each of the towns within the Dihlabeng

Municipality under the current climate (1961 to 1990). ........................................................ 45

Figure 20: The number of very high fire-danger days for each of the towns within the Dihlabeng

Municipality under the climate projected for the near future (2021 to 2050). ........................ 45

Figure 21: The number of very high fire-danger days for each of the towns within the Dihlabeng

Municipality under the climate projected for the far future (2070 to 2099)............................ 45

LIST OF TABLES

Table 1: Summary of the fire ecology types, the role of fire in ecosystem regeneration and the

fire characteristics (after Le Maitre et al., 2014b). The threshold of 650 mm of mean annual

rainfall (MAP) is based on studies summarised by Bond (1997) but this value is a guide rather

than being precise with low rainfall being associated with a low occurrence of fires (Archibald

et al., 2009) even in vegetation with grass fuels. ................................................................. 10

Table 2: The eleven natural land-cover classes with a description of each class (GTI, 2015)

........................................................................................................................................... 20

6

Table 3: The relationship between the natural land-cover classes (GTI, 2015) and the

likelihood of a fire (after Forsyth et al., 2010) ...................................................................... 21

Table 4: The relationship between the urban cover classes (GTI, 2015) and the likelihood of

a fire (after Forsyth et al., 2010). ......................................................................................... 23

Table 5: The relationship between the land-cover class for cultivated lands and the likelihood

of a fire. See Table 9 for a description of the Likelihood. Low, Medium and High describe the

vigour of the crop. ............................................................................................................... 25

Table 6: Descriptions of the social and economic consequence classes used in this

assessment (based on Forsyth et al., 2010). ....................................................................... 27

Table 7: The social and economic consequences based on the land-cover in the interior 1-km

wide buffer of the settlements (see Table 12 for descriptions of the consequence classes). 28

Table 8: The full risk matrix for assessing the social and economic risk that each settlement is

exposed to, (based on Forsyth, Kruger & Le Maitre 2010). ................................................. 30

Table 9: The dominant fire ecology type and corresponding fire sensitivity of the original and

remaining natural vegetation within the 1 km wide outside buffer on the WUI of each

settlement. The values are the percentages of the totals in each column. ........................... 33

Table 10: Towns categorised by the likelihood of a wildfire occurring within the Wildland-Urban

Interface .............................................................................................................................. 34

Table 11: Towns categorised by the economic consequences of a wildfire occurring within the

outer 1 km buffer of the WUI. For descriptions of the consequence classes see Table 6 .... 35

Table 12: The numbers of fires recorded (i.e. the number of times a MODIS pixel was detected

as burnt) in the outside buffer of each settlement. ............................................................... 46

Table 13: The percentage of area of the outside buffer burnt over a 15 year time period for

each settlement in each fire ecology type. ........................................................................... 47

7

1 INTRODUCTION

Most of the natural vegetation in South Africa is adapted to fires and requires fires to maintain

ecosystem function and biodiversity (Bond, 1997; Van Wilgen et al., 1990). At the same time

fires are a threat to human lives, livelihoods and assets and large losses are incurred in fires

every year1 by both the public and private sector. This tension – that fires are necessary but

fires are a hazard – is behind the approach taken to fires in natural vegetation in policy and

legislation, which supports the formation of Fire Protection Associations to manage fires and

prevent veldfires2. Climate change is projected to increase air temperatures, and increase the

periods between rainfall events, which are likely to increase the periods of high to extreme

fire-danger days and, thus, the risk of veldfires.

However, climate change will also alter vegetation productivity and, possibly structure, and

thus fuel dynamics which may influence the fire behaviour as well (Archibald et al., 2018; Bond

and Scott, 2010). The current study has not attempted to assess the impacts of climate change

on fuel dynamics because the development of suitable models for doing this for southern

African conditions is still an emerging field of research, so fuel dynamics are assumed to be

similar to what they are now.

Fire risk has been assessed at a national level for South Africa using information on the fire

hazards related to the fire regimes in different vegetation fire ecology types and the

consequences for human lives, livelihoods and assets (Forsyth et al., 2010). The 2010

assessment followed the well-established approach of dividing risk assessments into the

hazards and the consequences of exposure to those hazards, also termed the hazard and the

vulnerability. This assessment builds on and updates the 2010 assessment, focusing on fire

risk at the local authority, settlement and within settlement levels.

1 http://www.fpasa.co.za/journals/sa-national-fire-statistics 2 A wildfire is a fire burning in natural vegetation or veld or modified vegetation and could be a managed or unmanaged (wild) fire.

8

There is evidence that communities living in the wildland-urban interface of South African

towns are becoming increasingly exposed to wildfires due to, for example, expanding town

boundaries (Holloway et al., 2010; Nel et al., 2014; Syphard et al., 2013). Under conditions of

climate change it is predicted that the frequency and severity of high and extreme fire-danger

events and, thus, the risks will increase, at least in the Fynbos Biome (Le Maitre et al., 2015;

Midgley et al., 2005). Investments in protecting settlements need to prioritise those identified

as having the greatest risk. Therefore the objective of this study was to determine the level of

risk prevailing in the wildland-urban interfaces of the settlements of South Africa. A secondary

objective was to determine how the current number of high and extreme fire-danger days that

these settlements experience is likely to change under different climate change scenarios.

2 BACKGROUND

The following sections describe the overall approach and the details of the final methodology

are given under Methodology (section 3).

2.1 Fire hazard

Fire hazard depends on the nature of the fuels available to burn and the typical weather

conditions during which fires occur, especially wild fires. As data on the fuel loads in natural

vegetation types were only available for a very limited set of sites and fire ecology types, a

different approach was adopted (Forsyth et al., 2010; Le Maitre et al., 2014b). South Africa

has a long history of fire ecological research which has resulted in a large body of data and

information on the fire ecology, behaviour and fire regimes in the major vegetation types of

South Africa, notably grasslands, savanna-woodlands and fynbos (e.g. Bond, 1997;

Goldammer and de Ronde, 2004; Kraaij and Wilgen, 2014; Tainton, 1999; Teie, 2009; Van

Wilgen and Scholes, 1997). The available information was synthesised to identify 13 fire

ecology types and the fire regimes and behaviour were described for each of these types

(Forsyth et al., 2010; Le Maitre et al., 2014b). Each of the 430 different vegetation types in the

national vegetation map (Mucina and Rutherford, 2006) were assigned to one of these 13 fire

ecology types, resulting in a map of South Africa showing the distribution of the fire ecology

types (Figure 1).

9

Figure 1: Fire ecology types and their distribution in South Africa, Lesotho and Swaziland (Bond, 1997; Forsyth et al.,

2010; Le Maitre et al., 2014b).

The 13 fire ecology types can be arranged into three main groups (Forsyth et al., 2010; Le

Maitre et al., 2014b):

Fire-dependent ones which require fires at the right intervals to regenerate the

vegetation

Fire-independent ones which may burn but do not require fire for regeneration

Fire-sensitive ones which are adversely affected by fires and can take many years to

recover

Fire hazard will not be an issue for settlements in fire ecology types where fire hazard is low

or fires occur only rarely because there generally is insufficient fuel. These include the Nama

Karoo, Succulent Karoo and some forms of Thicket. Indigenous Forests only occupy very

limited areas so there are unlikely to be many, if any, settlements which are located in Forest

areas but if they did the fire hazard would be low. In Sparse Arid Woodland and in low rainfall

areas of Sweet Grassland, the hazard will vary depending on the rainfall and grass production

during the growing season (Table 1).

10

Table 1: Summary of the fire ecology types, the role of fire in ecosystem regeneration and the fire characteristics (after

Le Maitre et al., 2014b). The threshold of 650 mm of mean annual rainfall (MAP) is based on studies summarised by

Bond (1997) but this value is a guide rather than being precise with low rainfall being associated with a low

occurrence of fires (Archibald et al., 2009) even in vegetation with grass fuels.

Fire ecology

type Role of fire

MAP

(mm/year) Fire characteristics

Moist

Woodland

Fire-

dependent >650

Sufficient fuel for fires every year; in dense woodlands

there may be too little grass to sustain fires; woody

overstorey can carry fire; fire hazard can be high to

extreme

Arid Woodland Fire-

dependent 400-650

Sufficient fuel for fires in high rainfall years; woody

overstorey plants rarely burn; fire hazard can be high

Sparse Arid

Woodland

Fire-

independent <400

Sufficient fuel for fires only after exceptional rains;

sparse woody overstorey which does not carry fire;

fire hazard can be high

Sour

Grassland

Fire-

dependent, >650

Sufficient fuel for fires every year; fire hazard can be

high to extreme

Sweet

Grassland

Fire-

dependent <650

Sufficient fuel for fires in high to exceptional rainfall

years; fire hazard can be high

Coastal

Grassland

Fire-

dependent >650

Sufficient fuel for fires every year ; mixed woodland

and grassland, all carry fire; fire hazard can be high to

extreme

Fynbos Fire-

dependent

Fine fuel and litter accumulates over time; can burn

after 4-8 years; fire hazard can be high to extreme

Renosterveld Fire-

dependent

Fine fuel and litter accumulates over time; can burn

after 3-4 years; fire hazard can be high to extreme

Thicket Fire-

independent

Fuel loads low, does not normally burn; some grassy

forms burn periodically; fire hazard can be high

Nama Karoo Fire-

independent

Insufficient fuel; some forms with high sweet grass

cover can burn after exceptional rainfall years; fire

hazard can be moderate to high

Grassy Nama

Karoo Fire-sensitive

Accumulates sufficient fuel in high rainfall years; fire

hazard can be high

Succulent

Karoo

Fire-

independent

Insufficient fuel; very rarely burns; fire hazard low

Natural Forest Fire-sensitive Fuel properties limit fires; can burn in exceptionally

hot and dry conditions; fire hazard low to moderate

2.2 Changes in land-cover and fire hazard

The fire ecology types described above are all classified as natural vegetation in the land-

cover data (GTI, 2015). In many areas the land-cover has been converted from natural

vegetation to other land use and land-cover types and some of these land-cover types modify

the fuel loads. The most obvious one is the establishment of forest plantations of introduced

tree species, notably pines, eucalypts and wattles. Plantations of these species have high fuel

loads and considerable resources are invested in attempting to ensure that such plantations

11

areas are protected from fire (Van Wilgen and Richardson, 2014, 2012). Since the trees

require 5-30 years to reach a harvestable size this requires substantial investments in fire

protection and exclusion every year. When fires burn plantations they can present substantial

fire hazards to infrastructure and any settlements adjacent to or within plantation areas

(Goldammer and de Ronde, 2004; Teie, 2009). Fires in plantations can be particularly

damaging if they occur in areas that have recently been clear-felled and there are lots of debris

(slash) and other large woody material left behind. Fires under these conditions can be severe,

with large woody debris burning intensely and for long periods of time (hours to days) (Keeley,

2009), doing considerable damage to soils by consuming the organic matter that binds the

soil particles, and generating water-repellent conditions (Le Maitre et al., 2014a; Scott, 2000,

1997, 1993; Shakesby and Doerr, 2006; Shakesby, 2011). These conditions can alter a

catchment’s hydrological responsiveness and lead to unexpected flooding.

A range of crops are grown in South Africa but only some produce large quantities of fuel and

can become a fire hazard. The annual crops generally produce little if any fuel except for maize

but even in the case of maize, the residue ranges from 2.8-5.8 tons/ha (Batidzirai et al., 2016).

This is comparable with, and even higher, than the dry biomass reported for high-rainfall,

sourveld grassland at 3.4 tons/ha but closer to the total grassland biomass of 5.6 tons/ha

(Everson et al., 1988), suggesting that burning maize lands can result in a high to extreme fire

hazard. Compared with grassland, sugar cane is another crop which can generate large

amounts of fuels but it is generally harvested at a stage where much of the crop is still green.

This still leaves a residue but the biomass is relatively low. Orchards and vineyards can burn

in fires but generally where there is sufficient fuel in the weedy growth or cover crop

underneath the vines or trees. The biomass of this secondary vegetation is generally low so

that the fire hazard is low. An inspection of the areas regularly burnt in fires shows that fires

in agricultural lands are a regular occurrence. If these fires spread into the adjacent natural

vegetation they could results in hazards like those estimated in this study (Table 1).

A wide range of introduced (alien) plant species have also become significant invaders of

natural vegetation in South Africa. Many of these species grow taller, larger, or become denser

than the natural vegetation and thus increase fuel loads (Jayiya et al., 2004; Van Wilgen and

Richardson, 1985). The greater fuel loads can increase the fire hazard by increasing fire

intensity as well as increasing fire severity and the damage done to seed banks and soils

(Bond et al., 1999; Le Maitre et al., 2014a; Van Wilgen and Scott, 2001). There are data

available on density and species composition of invasions but these are mapped in a

12

generalised way (Kotzé et al., 2010). This means that the data are suitable for a general

overview of the invasions in an area but cannot give accurate data on invasions in specific

locations. In this study we have used the land-cover classes to infer the density of woody

plants, including invading alien shrub and tree species, and increase the fuel load and fire

hazard because the land-cover includes such invasions (without explicitly identifying them)

and provides greater spatial accuracy.

2.3 Fire vulnerability

Assessing fire vulnerability involved (Forsyth et al., 2010):

Identifying the conditions that increase the likelihood of loss for particular elements-at-

risk including: environmental (e.g. ecologically sensitive ecosystems), social (e.g.

human lives, livelihoods) and economic (e.g. infrastructure, built structures, agricultural

crops) elements that are exposed to a hazard, and are at risk of loss.

Determining the vulnerability level for different situations and conditions.

This study has focused on the social and economic risks. Forest plantations can be

environmentally sensitive because they can experience severe fires at times when high

ground fuel loadings (e.g. slash from thinning or felling) is present. These conditions result in

increases in storm runoff and soil erosion and changes in catchment properties (Scott et al.,

1998). Such changes were not included in this assessment.

Social impacts were assessed as those that directly affected life expectations and human

health and directly related to the ability of those social systems to recover (Forsyth et al.,

2010). A key factor in this is the financial reserves (e.g. savings) and the skills and equipment

required to protect against fires. This is directly related to financial income so areas with low

income were assessed as being the most vulnerable. Suitable income data are available for

±7 x 7 km spatial analysis units known as mesozones (Naudé et al., 2007; Van Huyssteen et

al., 2009) and the settlement typologies developed for this study. Social impacts can also be

inferred from the land-cover classes for urban settlements. Informal settlements around towns

and dense rural settlements are typically the most vulnerable because the entire settlement,

and all the household possessions, can be destroyed by a fire (a social and an economic loss).

13

Annual dryland crops were excluded from the economic impact assessment as they rarely

burn before harvest while irrigated crops generally do not burn (Forsyth et al., 2010). Perennial

crops, such as orchards, and forest plantations were included because re-establishment and

recovery after a fire may take a number of years. Harvestable natural products are difficult to

quantify and there are no maps so, although they are critical for many rural communities, they

could not be included. Livestock losses are often incurred during fires and should be

accounted for as well, with non-commercial farmers often being the most vulnerable, but

suitable datasets are lacking. Other economic losses would include those involving

infrastructure (e.g. power lines) and structures and facilities (e.g. houses, barns, factories,

clinics) which were included.

2.4 Fire risk

The potential for economic and social losses was rated based on a plausible wildfire scenarios

developed for each fire-ecotype (Forsyth et al., 2010). The fire scenarios were used to

evaluate the fire hazard and determine the likelihood and consequences of (i.e. vulnerability

to) such a fire in workshops with stakeholders. The combinations of likelihood – rated from

‘rare’ to ‘almost certain’ - and fire consequences – rated from ‘insignificant’ to ‘catastrophic’ –

were used to arrive at a level of fire risk from ‘low’ to ‘extreme’. In the 2010 study, the outcome

of the analysis was an assessment of the economic and social fire risk in each spatial unit

(mesozone) which was then aggregated to the level of the local authority to identify those most

at risk from fires. In this analysis the overall assessment was for all selected settlements in a

local authority, individual settlements and within individual settlements. Changes in the fire

risk in future can be accommodated by adjusting either the fire scenarios or the likelihood or

both. We do not plan to run stakeholder workshops to redo the fire risk assessment process

and, so, will rely on the findings of the 2010 assessment for the actual ratings of fire risk.

14

2.5 Modelling fire danger

The fire ecology types can be divided into groups based on the likelihood of the occurrence of

a wildfire (Table 1). The Succulent Karoo, Nama Karoo and Thicket rarely burn because they

have little or no fuel except for some grassy forms of Thicket and thicket-grassland transitions

with higher grass cover. Sparse Arid woodland and Sweet Grassland will burn only after

exceptionally high rainfall years result in sufficient grass fuels. Fires in Fynbos and

Renosterveld occur regularly with their occurrence determined by cumulative fuel loads, which

usually take 4-6 years to be sufficient to carry a fire. The other savanna and grassland types

can accumulate sufficient fuel burn every year but mostly burn every two years (Le Maitre et

al., 2014b). The Forest rarely burns and can be considered a low fire hazard for this analysis.

This means that the focus needs to be on modelling the fire hazard in two groups of fire

ecology types:

Fynbos and Renosterveld with a mix of woody and herbaceous fuels

Moist and Arid Woodland and Sweet, Sour and Coastal Grassland where grass fuels

are dominant

2.5.1 Fynbos and Renosterveld

Fynbos and Renosterveld occur primarily in the winter rainfall region and the dry summers

result in dry fuels and ideal conditions for fires, but both Fynbos and Renosterveld extend

eastwards to near Port Elizabeth and grassy forms of Fynbos occur in the Zuurberg and the

Kaprivierberge east of Grahamstown. Here the rainfall is bimodal with a mixture of summer

and winter and the dry conditions and fires can occur throughout the year (Kraaij et al., 2013).

For the Fynbos the fuels comprise a mixture of fine-leaved shrub and grass like fuels such as

restios, sedges and grasses, except where there are dense stands of Proteaceae (Le Maitre,

2015). Decomposition rates are slow so that fine dead fuels accumulate and play an important

role in fire ignition and propagation. Renosterveld lacks the Proteaceae, restios are rare or

absent and some forms have abundant geophytes and annuals during the winter which add

to dry fuel load in the summer.

Fire danger in these fire ecology types is well captured by the McArthur forest Fire Danger

Index (FDI) and less well by the Lowveld FDI (Willis et al., 2001). Algorithms were developed

for calculating these indexes from the climate data outputs from the CSIR global climate

modelling system (Forsyth et al., 2015). The algorithms were used to calculate the values of

15

the FDIs for both current and future climates. The McArthur forest FDI can also be applied to

the rest of the country but is not best suited to grass-dominated fuels as discussed below.

2.5.2 Woodland and Grassland

The onset of grass growth is triggered by the increase in soil moisture following the onset of

the rainy seasons in the summer rainfall region (Archibald and Scholes, 2007; Parrini and

Erasmus, 2009; Whitecross et al., 2017). In the fire ecology types with grass dominated fuels,

the seasonality of the grasses is important. In the high altitude and Highveld grasslands the

seasonality is determined by minimum temperatures which determine when grasses begin

growing and very low temperatures, especially frosts which kill the grass above the soil (Ellery

et al., 1995). In the drier areas the grass growth is regulated primarily by moisture availability

which determines the onset of the growing season, and moisture stress linked to the onset of

the dry season. Here the amount of rainfall during the rainy season plays two roles, regulating

the amount of growth (Deshmukh, 1984; Govender et al., 2006; Hély et al., 2003; Snyman

and Fouché, 1991) and the onset of moisture stress in conjunction with temperature (Ellery et

al., 1995).

The process of grass die-back is known as curing. Curing is important because it determines

the amount of live versus dry fuel and thus the fuel mix and load and conjunction with rainfall,

and thus the resulting fire behaviour under a given set of conditions (Archibald et al., 2010a;

Govender et al., 2006). The dynamics of grass fuels play an important role in the extent of

fires (e.g. burned area) and the seasonality of fires, with Sour Grasslands in the interior of

KwaZulu-Natal burning far more extensively and frequently than the coastal belt grasslands

and the valley Bushveld (Buthelezi et al., 2016) (Figure 2). The interior grasslands also have

a more strongly defined burning season than the coastal grassland and Valley Bushveld

(Figure 2). Given that the curing is dynamic and driven by temperature or moisture availability,

it would be ideal to be able to simulate it for future climates from data on the key variables

rather than have a static start and end date. Although there have been studies of the

phenology of the grasses-fuelled biomes (savanna, grasslands) in South Africa and how that

phenology varies with annual rainfall, these models focused on the timing rather than the cues.

The following rules were suggested for modelling the phenology of Highland Sourveld which

is found in the Drakensberg and its foothills (Everson et al., 1985): Curing was initiated by the

following conditions, a minimum temperature of 0°C for 5 successive days or 1 day of -4°C.

All live fuel would then be moved to dead fuel. New growth or greening up would take about

16

14 days and have a high moisture content (250%) before decreasing to 120% in mature foliage

(equations provided). The same rules could probably be applied to KwaZulu-Natal Sandstone

Sourveld, possibly to Ngongoni Veld but not to the coastal grasslands. The phenology of South

African vegetation has been studied and divided into biome related groups (Hoare and Frost,

2004; Wessels et al., 2011). The moister savannas were found to be distinct from the arid

savannas, mainly because the arid savannas respond to inter-annual rainfall variations and

lack the clear peaks of the moister savannas. The Indian Ocean Coastal belt was separated

from the remaining grasslands but displays a similar winter dormancy period to the grassland

and savanna biomes (Wessels et al., 2011) which fit with the seasonality of fires (Figure 3).

For practical reasons the McArthur Forest FDI was used in this study for all the fire ecology

types.

Figure 2: Total burnt area (expressed as proportion of total area within each vegetation type) in the (a) KZN Sandstone

Sourveld, (b) Ngongoni Veld, (c) Eastern Valley Bushveld and (d) KZN Coastal Belt over a 10-year period (from

Buthelezi et al., 2016).

17

Figure 3: Percentage of the KwaZulu-Natal Sandstone Sourveld, Ngongoni Veld, Eastern Valley Bushveld and

KwaZulu-Natal Coastal Belt that burnt in each month (March to September) (from Buthelezi et al., 2016).

3 METHODOLOGY

In this study we adapted the process we followed in the 2010 study (Forsyth et al., 2010),

which used the mesozones (irregular polygons about 50 km2 in area) as the basic assessment

unit to use the settlement boundaries as the assessment unit. We added two components to

this assessment, namely fire occurrence information supplied by the CSIR’s Meraka Institute

and based on MODIS data, and the projected changes in the occurrence of fire-danger days

generated by the Green Book project.

3.1 Defining the area for assessing the hazard

The settlement boundaries or footprints were supplied by the Green Book project as a set of

named polygons to which a number of attributes were linked. We focused on the outer

boundaries of these settlements as defining the centre of the Wildland-Urban Interface (WUI)

where urban or other infrastructural development borders on and intermingles with flammable

natural vegetation, called wildlands in the United States of America (Theobald and Romme

2007). The WUI is the area where residents and their assets are exposed to death or injury

and damage from wildland fires. Assets are things that society values and fire-sensitive assets

can be social, economic and environmental in nature (Gill et al., 2013). Examples of social

18

assets include human lives and homes, businesses, houses, schools and sheds, while

environmental assets could include wetlands or biodiversity.

There is no universal standard for the dimensions of the WUI and different studies have

adopted a variety of approaches including set widths and densities of structures (Conedera et

al., 2015; Ghermandi et al., 2016; Herrero-Corral et al., 2012; Lampin-Maillet et al., 2010;

Whitman et al., 2013), often depending on the purpose and spatial resolution of the study.

For a very detailed, aerial photograph-based assessment of the WUI in two areas of South

Africa we chose to use a width of 100 m on either side of the wildland-urban boundary for the

risk assessment (Forsyth and Le Maitre, 2015). For this national level assessment we chose

the commonly used width of 1 km. This is appropriate given that our primary concern was the

impact on infrastructure, particularly houses which are susceptible to being set on fire during

high fire danger conditions by wind-borne embers carried for at least 1 km by wind before they

land in gutters, enter roof eaves or open windows and ignite fires (Alexandre et al., 2016;

Blanchi et al., 2006).

A number of the settlements comprised a number of sub-units which adjoined each other on

one or more sides, so the first step was to dissolve them so that there was a single, continuous

boundary to use in defining the inside and outside. That polygon boundary was then buffered

using ArcGIS to a width of 1 km on both the inside and the outside. The outside or exterior

buffer was used to analyse the land-cover classes to estimate the fuel load and fire danger

(likelihood) and the inside or interior buffer to assess the consequences of, or vulnerability to,

a wildfire. So, the WUI was defined as a 2 km wide strip with its centre defined by the boundary

of the settlement provided by Workstream #3. We know from inspection of the land-cover data

that there can be natural vegetation on the inside and/or infrastructure on the outside of the

centre line, but using this boundary would make the findings comparable across all the

different studies in this project and for all settlements. In many cases the centre of the

settlement is less than 1 km from the edge so the entire settlement is in the inner buffer.

3.2 Characterising fuels and fire hazard

There are various ways of estimating the fuel loads, ranging from vegetation or land-cover

classifications to field sampling. From previous studies we know that only certain vegetation

types in South Africa burn regularly in fires and thus pose a hazard and these have been

grouped into a set of fire ecology types (Le Maitre et al., 2014b). However, this only gives a

19

broad distinction between fire ecology types that burn regularly and those that rarely burn or

have never had a detectable fire (Archibald et al., 2010b, 2009; Le Maitre et al., 2014b). Finer

resolution data are needed to provide information on the structure and other characteristics of

the vegetation that influence fire behaviour and, thus, the fire hazard (Baeza et al., 2002;

Everson et al., 1985; Van Wilgen, 1984).

The National Land-Cover of 2013-14 provides a number of classes of natural vegetation and

transformed land-cover classes that can be used to infer the fuel structure and characteristics

(GTI, 2015). The classification was automatically generated from remote sensing images

using an algorithm which analysed changes over a time series of images which represented

a full seasonal cycle. The data are available at a spatial resolution of 30x30 m. There are 72

classes in the land-cover dataset which can be divided into the following broad categories:

11 for natural or degraded natural land-cover types, which include one for permanent

water bodies such as lakes and man-made dams

22 for cultivated land under various crops

for commercial forest plantations and woodlots

for mines and mine infrastructure

29 for urban areas which also include information on the vegetation within the urban

matrix and indicate the size of the land holdings

The natural classes give an indication of the vegetation density and the structure and were

used as the basis for assigning the Fire Hazard in the form of the ‘Likelihood of a wildfire’

(Forsyth et al., 2010).

20

Table 2: The eleven natural land-cover classes with a description of each class (GTI, 2015)

Class Name Description

Water seasonal

Areas of open, surface water that is only detectable in certain

seasons; typically vleis, pans or small dams that dry out; seasonal

vegetation and reeds can fuel fires and be a fire hazard; often very

limited in extent

Water permanent

Areas of open, surface water detectable all year round; includes

both natural (e.g. lake) and man-made water features (e.g. dams);

mapped extent represents the maximum detectable water extent

during the 2013-14 assessment period

Wetlands

Wetland areas with a high vegetation cover seasonally or all year

round; includes seeps/springs, marshes, floodplains, lakes / pans,

swamps, estuaries, and some riparian areas and such area within

areas under subsistence cultivation; typically have more biomass

and, thus, potentially more fuel than adjacent vegetation and a fire

hazard

Indigenous Forest Natural or semi-natural forest, dominated by tall trees (>5 m),

canopy cover >75%, often with understory vegetation canopies

Thicket /Dense bush

Natural or semi-natural tree and/or bush (shrub) dominated areas,

canopy heights 2-5 m, and canopy cover >75%, can include

localised lower cover areas (±60%); includes a range of woody

vegetation types

Woodland/Open bush

Natural or semi-natural tree and/or bush (shrub) dominated areas,

canopy heights 2-5 m and canopy density 40-75%, can include

localised lower cover areas (15-20%); includes a wide range of

bush and woodland vegetation types, often with a high ground

cover of grasses

Grassland

Natural or semi-natural grass dominated areas, where woody plant

cover is <20%, can include localised higher cover (<40%) such as

bush clumps; may include planted dryland pastures

Shrubland fynbos

Natural or semi-natural shrub dominated areas, canopy height

typically <2 m, mainly associated with the Fynbos Biome; canopy

cover very variable from low to high; low cover is normal for a few

year after fires and in very rocky areas

Low shrubland

Natural or semi-natural shrub dominated areas, canopy height

typically <2 m; mainly in arid areas including the Karoo; canopy

cover very variable from low to high; low cover is normal in very

rocky areas and degraded areas

Erosion (donga) Areas with dongas mapped in various studies by GTI and others;

probably under-estimated in well vegetated areas

21

Class Name Description

Bare none vegetated

Areas of bare ground (soil) or sparse to very sparse vegetation

cover (<5-10%), either of natural origin or due to natural (e.g. soil

slip) or man-induced processes (e.g. advanced vegetation

degradation and erosion), bare rock, dry river beds, dry pans,

coastal dunes and beaches; inland sand dune and sandy areas;

may include recent fire scars in fynbos areas

There is a clear increase in biomass from low shrubland to woodland or thicket to forest but

this does not directly translate into an increased fuel load, and thus fire hazard. In the case of

forest the structure of the vegetation and the fuel characteristics tend to make forests fire-

resistant (Van Wilgen et al., 1990). However during dry and warm to hot weather and strong

winds, fires can burn and spread through forests (Geldenhuys, 1994; Phillips, 1931),

especially the drier forms of forest (as seen in the recent Knysna fires). Nevertheless, the

likelihood of fires in forests is considered ‘Rare’ (Error! Reference source not found.). The

oody biomass in thicket and woodland does not necessarily contribute to fuel loads, especially

in savanna, where the primary fuel is the grass layer. However, the grass layer in areas with

a rainfall of at least 600-650 mm accumulates enough fuel to sustain fires every year and so

the likelihood is assessed as ‘Likely’ (Bond, 1997; Bond et al., 2003; Le Maitre et al., 2014b).

Many forms of thicket do not burn but some forms do, while shrubland fynbos does burn giving

them both a likelihood of ‘Possible’.

Table 3: The relationship between the natural land-cover classes (GTI, 2015) and the likelihood of a fire (after Forsyth

et al., 2010)

Class Name Likelihood Description

Water seasonal Possible Likelihood of a fire is once in 10 years;

might occur at some time; as likely as not

Water permanent Rare

Likelihood of a fire is once in 100 years;

may only occur in exceptional

circumstances

Wetlands Possible Likelihood of a fire is once in 10 years;

might occur at some time; as likely as not

Indigenous Forest Rare

Likelihood of a fire is once in 100 years;

may only occur in exceptional

circumstances

Thicket /Dense bush Possible Likelihood of a fire is once in 10 years;

might occur at some time; as likely as not

Woodland/Open bush Likely Likelihood of a fire is once in 5 years; will

probably occur

22

Class Name Likelihood Description

Grassland Likely Likelihood of a fire is once in 5 years; will

probably occur

Shrubland fynbos Possible Likelihood of a fire is once in 10 years;

might occur at some time; as likely as not

Low shrubland Rare

Likelihood of a fire is once in 100 years;

may only occur in exceptional

circumstances

This classification was then used to assess the likelihood of a fire in other land-cover classes,

especially the urban classes which often included a vegetation component (Table 4).

Commercial forest plantations can have high fuel loads and are often of inherently flammable

species, notably eucalypts and pines with their high wax and resin contents which tend to lead

to high intensity fires that are difficult to control (Gill and Zylstra, 2005; Schwilk, 2015). These

species are well adapted to fires in their native habitats and fires trigger spread and seedling

recruitment, increasing the density and this is also what makes the pines, in particular, very

successful invaders in many parts of the world (McConnachie et al., 2015; Richardson et al.,

2000; Richardson and Bond, 1985). The increased fuel loads in areas invaded by these same

tree species, and many other tree and shrub species, result in fires that are difficult to control

and they contribute to the fire hazard in South Africa. Invasions by these species are not

shown in the land-cover as such, but can be inferred from information on the tree and shrub

cover in the descriptions when compared with the natural vegetation for those areas. We have

assumed that high woody plant cover is likely to reflect some degree of alien plant invasion in

those areas and have allowed for that in this assessment. The land-cover classification

currently recognises three stages of plantations and/or woodlots: mature trees, young trees

and clearfelled. Since the stage of plantation areas varies over time as they grow, mature and

are clearfelled, the characteristics will change over time and all plantation/woodlot areas could

be treated as one likelihood. We decided to distinguish between the likelihoods in the different

stages to illustrate the importance of periodic reassessments of the fire hazard as part of the

ongoing fuel management. So, young and clearfelled plantations/woodlots with their relatively

low fuel loads are given a likelihood of ‘Possible’ while mature plantations/woodlots with higher

fuels are rated ‘Likely’. Clearfelled areas do have high fuel loads from the unharvested

material, but this is typically close to the ground and this stage typically is only present for a

short period while the plantation is re-established so we have not increased the likelihood.

The urban classes are divided into broad groups based on the density, the mix of building

types and degree of formal planning and then also by the structure of the vegetation (Table

23

4). We have also used our experience based on an assessment of the fire hazard on the WUI

in the Helderberg area near Cape Town (Forsyth and Le Maitre, 2015) and the impacts of a

subsequent wildfire in this area (G. Forsyth unpublished data) to assess the likelihoods for

different urban classes.

Table 4: The relationship between the urban cover classes (GTI, 2015) and the likelihood of a fire (after Forsyth et al.,

2010).

Class Name Characteristics Likelihood

Urban built-up: dense or

open trees/bush

Well-established, usually planned,

moderate density, mixed use; built-up with

some natural vegetation: gardens and

public open space; bush can be an

important fuel and so can the ground layer

Likely

Urban informal: dense or

open trees/bush

Unplanned, high density development

interspersed with tree/bush cover; bush

can be an important fuel and so can the

ground layer

Likely

Urban residential: dense or

open trees/bush

Well-established, usually planned,

moderate density, residential; largely built-

up with natural vegetation, gardens and

public open space; bush can be an

important fuel and so can the ground layer

Likely

Urban smallholding: dense

or open trees/bush

Low density mix of buildings & other built-

up structures, may include open and

cultivated areas; formally declared

agricultural holdings, and similar small

holdings/small farms, typically located on

the periphery of urban areas; extensive

bush areas can be an important fuel bed

and so can the ground layer

Likely

Urban sports and golf:

dense or open trees/bush

Low density mix of buildings & other built-

up structures associated with golf courses;

typically represents the border extent of

the entire estate or course; bush and less

tended areas can accumulate fuels

Likely

Urban township: dense or

open trees/bush

High density buildings and other built-up

structures; formal, regulated, residential

(RDP) housing; bush cover limited but

usually untended and accumulates fuels

Likely

24

Class Name Characteristics Likelihood

Urban village: dense or

open trees/bush

Variable density of structures; typically

associated with rural villages, modern and

traditional buildings; bush and less tended

areas can accumulate fuels

Likely

Urban built-up: low

vegetation/grass

As for built-up above, low fuel loads but

grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban informal: low

vegetation/grass

As for informal above, low fuel loads but

grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban residential: low

vegetation/grass

As for residential above, low fuel loads but

grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban smallholding: low

vegetation/grass

As for smallholding above, low fuel loads

but grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban sports and golf: low

vegetation/grass

As for sports and golf above, low fuel

loads but grass fuels can accumulate

potentially raising the Likelihood

Possible

Urban township: low

vegetation/grass

As for township above, low fuel loads but

grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban village: low

vegetation/grass

As for village above, low fuel loads but

grass fuels can accumulate potentially

raising the Likelihood

Possible

Urban built-up (bare) As for built-up above, little or no fuel Rare

Urban informal (bare) As for informal above, little or no fuel Rare

Urban residential (bare) As for residential above, little or no fuel Rare

Urban smallholding (bare) As for smallholding above, little or no fuel Rare

Urban sports and golf

(bare)

As for sports and golf above, little or no

fuel Rare

Urban township (bare) As for township above, little or no fuel Rare

Urban village (bare) As for village above, little or no fuel Rare

Urban school and sports

ground

Open areas with mown grass cover, little

or no fuel Rare

Urban commercial

Formal, variable density and size of

structures, often with extensive paved

areas, little or no fuel

Rare

Urban industrial

Formal, variable density of generally large

structures, often with extensive paved

areas, little or no fuel

Rare

25

The next group was the cultivated land-cover classes where the primary distinction was

between perennial crops (e.g. orchards) and annual crops (e.g. small grains, maize), and

between irrigated crops and dryland crops (GTI, 2015) (Table 5). It is important to distinguish

between commercial farming where the fields are interspersed with limited areas of natural,

modified or degraded natural vegetation and subsistence cropping typically with small fields

forming a mosaic with more extensive modified or degraded natural vegetation which would

influence fire behaviour. For this assessment we have not increased the likelihood for

subsistence cropping because the natural vegetation mosaic will have its own likelihood, but

it could be considered in future. The land-cover also distinguished between crop vigour

classes based on the Normalised Difference Vegetation Index with higher crop vigour

indicating higher productivity and thus the potential for greater fuel load. We have not used

this information to modify the likelihood class at this stage. For sugar cane, no distinction was

made between commercial farming and emerging farmers because the crop is essentially the

same.

Table 5: The relationship between the land-cover class for cultivated lands and the likelihood of a fire. See Table 9 for

a description of the Likelihood. Low, Medium and High describe the vigour of the crop.

Class Name Characteristics Likelihood

Cultivated commercial fields (low,

medium, high)

Annual crops, large fields and farms,

little natural vegetation Possible

Cultivated subsistence (low,

medium, high)

Annual crops, small fields and farms,

mosaic with natural vegetation Possible

Cultivated commercial pivots (low,

medium, high)

Annual crops, high intensity cropping

of vegetables and other crops, little

fuel

Unlikely

Cultivated cane commercial &

emerging farmer’s – crop fields or

pivots

Perennial crop, cane can burn

seasonally, commercial farmers

have larger fields and farms than

emerging farmers

Likely

Cultivated cane commercial &

emerging farmers – fallow fields or

pivots

Perennial crop, cane can burn

seasonally, fuel load in fallow stage

could be greater but fields on a cycle

so the crop/fallow is not constant;

Possible

Cultivated vines (low, medium &

high)

Perennial crop, vines and cover

crops can burn Possible

Cultivated orchards (low, medium,

high)

Perennial crop, low fuel and green

cover crops Unlikely

26

The final step was to use the Zonal Statistics function in ArcGIS to calculate the area of each

of the land-cover classes within the 1 km wide exterior buffer so that we could estimate the

likelihood and, thus, the fire hazard in this portion of the WUI for each settlement.

3.3 Quantifying the fire hazard for each settlement

The output from the fire hazard (likelihood) assessment above was a GIS dataset with the

area of the 1 km exterior buffer in each of the land-cover classes in that buffer. The land-cover

class information was then joined to the corresponding likelihood class in the GIS to create a

set of likelihoods matched to each land-cover class. The database was then exported to Excel

as two datasets: settlements beginning with the letters A-L and those beginning with M-Z so

that the numbers of records fitted within the limitation on the numbers of records that Excel

can manage. A pivot table was then generated to calculate the areas in each likelihood class

in each settlement and that was then used to calculate the proportion of each likelihood class

in each settlement. Then the likelihood class with the highest proportion and >25% in each

settlement was selected as the final class for each settlement. The rationale for using >25%

was that there were four classes and a maximum of 25% would indicate an even split. If the

split was even then the highest likelihood class would be chosen but in practice this did not

happen. The final likelihood class was then joined to the settlement layer to create a spatial

dataset with the likelihood and thus the fire hazard for each settlement.

3.4 Determining the consequences of a wildfire for each settlement

We followed essentially the same steps as we had with the fire hazard but this time with the

land-cover classes present in the interior buffer. The first step in this process was to calculate

the area in each land-cover class within the internal buffer as described above for the fire

hazard.

In parallel, we used the characteristics of the land-cover in each settlement to generate ratings

for social and economic consequences for each land-cover class in the event of a wildfire. The

social consequences were assessed using a scale developed previously (Forsyth et al., 2010)

which rates the impacts on the social assets of human life and health based on exposure to

the hazard of wildfires (Table 6).

27

Table 6: Descriptions of the social and economic consequence classes used in this assessment (based on Forsyth et

al., 2010).

Level of Consequence

Consequence end points

Social consequence

criteria

Economic consequence

criteria

Catastrophic

Death of one or more

persons in the scenario,

probable smoke inhalation

and other health impacts;

probably requiring

emergency hospitalisation

and affecting work capacity;

or, evacuation required

Depressed municipal

economy; extensive and

widespread loss of assets;

major impact across a large

part of the community; long-

term external assistance

required to recover

Major

Extensive injuries to people

in the scenario, probable

smoke inhalation and other

health impacts; probably

requiring emergency

hospitalisation and affecting

work capacity; or,

evacuation required

Serious financial loss,

affecting a significant portion

of the community; those

affected require external

funding (e.g. from Disaster

Management funds) to

recover

Moderate

Medical treatment required

but full recovery possible,

possible smoke inhalation

and other health impacts;

possibly requiring

emergency hospitalisation

Localised damage to

property and some financial

loss; short-term external

assistance required to

recover

Minor

Minor injuries only, possible

smoke inhalation and other

health impacts, may require

first aid treatment

Minor financial loss; short-

term damage to individual

assets; no external

assistance required to

recover

Insignificant No injuries, no other health

impacts, no first aid required

Inconsequential or no

damage to property;

negligible losses

Social vulnerability is generally directly related to people’s wealth which, in turn can be linked

to the kinds of housing they occupy and the services within their reach. Another dimension of

the social consequences, which is also linked to the economic consequences, is the likelihood

of the loss of dwellings and the insecurity that goes with being moved to, and having to live in

temporary accommodation and the accompanying social disruptions and stresses (Edgeley

and Paveglio, 2017; Holloway et al., 2010). Then there is also the potential for reduced health

and disablement due to, for example burns or smoke inhalation, and even death and disruption

of family life.

28

Table 7: The social and economic consequences based on the land-cover in the interior 1-km wide buffer of the

settlements (see Table 12 for descriptions of the consequence classes).

Land-cover Class Name1 Social

Consequences Economic

Consequences

Water seasonal Minor Insignificant

Water permanent Insignificant Insignificant

Wetlands Minor Minor

Indigenous Forest Insignificant Minor

Thicket /Dense bush Major Minor

Woodland/Open bush Major Moderate

Grassland Minor Moderate

Shrubland fynbos Moderate Moderate

Low shrubland Minor Minor

Cultivated commercial fields (low, medium, high) Minor Minor

Cultivated commercial pivots (low, medium, high) Minor Minor

Cultivated orchards (medium, high) Minor Moderate

Cultivated orchards (low) Minor Minor

Cultivated vines (low, medium, high) Minor Moderate

Cultivated permanent pineapple Minor Minor

Cultivated subsistence (low, medium, high) Minor Moderate

Cultivated cane pivot - crop Minor Moderate

Cultivated cane pivot - fallow Minor Minor

Cultivated cane commercial emerging – crop or fallow

Minor Minor

Plantations / Woodlots mature Moderate Major

Plantation / Woodlots young Minor Moderate

Plantation / Woodlots clearfelled Moderate Minor

Mines bare, semi-bare, water seasonal or permanent

Insignificant Insignificant

Mine buildings Minor Minor

Erosion (donga) or Bare none vegetated Insignificant Insignificant

Urban commercial or industrial Minor Major

Urban informal (dense trees / bush) Major Catastrophic

Urban informal (open trees / bush or low veg / grass)

Moderate Major

Urban informal, residential or smallholding (bare) Insignificant Insignificant

Urban residential (dense trees / bush) Moderate Moderate

Urban residential (open trees / bush or low veg / grass)

Minor Minor

29

Land-cover Class Name1 Social

Consequences Economic

Consequences

Urban school and sports ground Minor Major

Urban smallholding (dense trees / bush) Moderate Major

Urban smallholding (open trees / bush or low veg / grass)

Minor Moderate

Urban sports and golf (dense tree / bush) Minor Moderate

Urban sports and golf (open tree / bush or low veg / grass)

Insignificant Minor

Urban sports and golf (bare) Insignificant Insignificant

Urban township (dense trees / bush) Moderate Major

Urban township (open trees / bush or low veg / grass)

Moderate Moderate

Urban township (bare) Insignificant Insignificant

Urban village (dense trees / bush) Moderate Major

Urban village (open trees / bush or low veg / grass)

Moderate Moderate

Urban village (bare) Insignificant Minor

Urban built-up (dense trees / bush) Moderate Major

Urban built-up (open trees / bush low veg / grass) Minor Moderate

Urban built-up (bare) Insignificant Insignificant

The economic effects of wildfires have generally not been thoroughly documented although

there are some studies and data on particular sectors such as agriculture and forestry (Van

Wilgen et al., 2010). For this assessment, the economic consequences were based on the

degree or magnitude of the social and economic consequences for the community. These

assets included infrastructure (such as power lines), industrial facilities (e.g. sawmills), fodder,

livestock, homesteads, resorts, and plantation forests. Dryland crops, such as maize, were

excluded since these seldom suffer significant losses due to wildfires. There may also be

losses of harvestable natural resources, such as thatching-grass which could have detrimental

impacts on local communities. In some areas, notably grasslands, there may be losses of the

rangeland vegetation that would have provided fodder for livestock and may require short-

term donations or subsidies but the focus here in on the settlements rather than the farms.

However, in many rural areas, the people in the settlements are also the owners of the

livestock and may experience economic losses or hardship.

3.5 Determining the final level of social and economic consequence

for each settlement

30

The output from the social and economic consequence assessment above was a GIS dataset

with the area in the 1 km interior buffer in each of the land-cover classes in that buffer. The

land-cover class information was joined to the corresponding social and economic

consequence class in the GIS to create a set of consequences matched to each land-cover

class. The database was then exported to Excel as two datasets: settlements beginning with

the letters A-L and those beginning with M-Z so that the numbers of records fitted within the

limitation on the numbers of records that Excel can manage. A pivot table was then generated

to calculate the areas in each social and economic consequence class in each settlement and

the proportion of each consequence class in each settlement was calculated. The social and

economic consequence class with the highest proportion and also >20% in each settlement

was selected as the final class for each settlement. There were five consequence classes

(Table 6) so a maximum of >20% rule was used because it indicates an even split across the

five classes. If the split had been even then the highest likelihood class would have been

chosen manually but in practice this did not happen. The final social and economic

consequence class was then joined to the settlement layer to create a spatial dataset with the

social and economic consequences for each settlement.

3.6 Using a risk matrix to assign a fire risk rating

The final step involved using a risk matrix or look-up table to assign a fire risk rating that took

both the final fire danger (likelihood) and the final social and economic consequences for each

town into account (Forsyth, Kruger & Le Maitre 2010) (Table 8).

Table 8: The full risk matrix for assessing the social and economic risk that each settlement is exposed to, (based on

Forsyth, Kruger & Le Maitre 2010).

Likelihood Rating

Consequence Rating

Insignificant Minor Moderate Major Catastrophic

Almost certain

Medium Medium High Extreme Extreme

Likely Low Medium High Extreme Extreme

Possible Low Medium High High Extreme

Unlikely Low Low Medium High Extreme

Rare Low Low Low Medium High

The result of this process was an assessment of the fire risk for every one of the roughly 1600

settlements examined in this study.

31

3.7 Accounting for the effects of climate change on fire danger

As described in section 2.4, the occurrence of fire-danger days was calculated using the

McArthur Forest fire danger index (FDI) which has been successfully tested in South Africa

(Willis et al., 2001). We used the algorithms developed for calculating the FDI from the climate

data outputs from the CSIR global climate modelling system (Forsyth et al., 2015) to calculate

the daily values of the fire-danger indexes for both current and future climates. The data on

fire-danger days were generated from the outputs of an ensemble of six global change models

run with different greenhouse gas emission scenarios. In each case we used the worst case

emission scenarios where there is little mitigation (RCP8.5). Each of the models was run for

three climate change periods: baseline (1961 – 1990), near future (2021 - 2050) and far future

(2070 -2099). The McArthur Forest Fire Danger Rating Model (Noble et al., 1980) algorithm

was run on each of the three time period datasets to calculate the annual number of very high

fire-danger days for an 8 x 8 kilometre grid covering southern Africa. The number of extreme

and high fire-danger days was then extracted for each settlement polygon for the current, near

and far future climates.

3.8 Using records for fire occurrence

In addition, we also used information on the occurrence of fires between 2001 and 2016 from

the MODIS burnt-area product produced for the Advanced Fire Information System by the

Meraka Institute. This summary gives the number of times a fire occurrence has been detected

in each MODIS pixel (500 m x 500 m) during that period. The data are not perfect as they

include both false detections or false positives (i.e. recording a fire where there wasn’t one)

and detection failures or false negatives (i.e. not detecting a fire) with an overall accuracy of

about 75% (Archibald et al., 2010a; Roy and Boschetti, 2009; Tsela et al., 2014). This dataset

was used to estimate the number of times a fire has occurred within the exterior buffer of each

settlement. This provides a cross-check on the fire hazard and fire risk classification of each

of the settlements.

4 RESULTS

4.1 Fire hazard or likelihood

32

Altogether, the likelihood and consequences of a wildfire occurring within the WUI has been

examined for 1 596 settlements (Table 16). Each of these settlements is located within a

particular fire ecology type, although some may be located on the boundary between one or

more fire ecology types. These fire ecology types were the original vegetation in that area but,

over time, parts or even all of the vegetation within the 1 km outside buffer have been modified

and transformed, some potentially being a greater fire hazard. For example, invasion of fynbos

or grassland by pines or eucalypts. However, for those that have some natural vegetation

remaining, this gives an indication of the inherent fire-hazard also posed by natural vegetation

outside that buffer. We analysed the relationship between the settlements and the original fire

ecology types and found that 84% of all the settlements fell into a fire-dependent fire ecology

type (Table 9). The sensitivity to fire indicates both that it would require fires to maintain

ecosystem structure and function and that it burns naturally given the right conditions and a

source of ignition (Le Maitre et al., 2014b) (Table 13). This also means that the local fire

protection association or authorities would need to actively use fires to maintain the vegetation

and manage fuel loads. About 48% of these settlements occur in Moist Woodland or Sour

Grassland where fires can occur every year and the grass fuel layer requires fires every two

to five years to rejuvenate it. About 14% of the settlements occur in ‘Fire Independent’

vegetation with most of them in the Nama Karoo, Thicket and Succulent Karoo fire ecology

types where fires are rare or absent. Only 2% of the settlements are found in ‘Fire Sensitive’

vegetation and just over 80% of them are in the Grassy Nama Karoo fire ecology type which

does occasionally burn, especially after high spring and summer rainfall. Sweet Grassland

and Arid Woodland fire ecology types also require high rainfall to produce sufficient grass fuel

for fires but when they do accumulate fuels, the fires can be very extensive and the authorities

need to be prepared for such fires. The outside buffer of one settlement, not shown in Table

9, was dominated by a water body.

33

Table 9: The dominant fire ecology type and corresponding fire sensitivity of the original and remaining natural

vegetation within the 1 km wide outside buffer on the WUI of each settlement. The values are the percentages of the

totals in each column.

Fire ecology type Fire sensitivity

Fire dependent Fire independent Fire sensitive

Arid Woodland 19.69 0.00 0.00

Coastal Grassland 4.46 0.00 0.00

Forest 0.00 0.00 19.23

Fynbos 11.14 0.00 0.00

Grassy Nama Karoo 0.00 0.00 80.77

Moist Woodland 20.21 0.00 0.00

Nama Karoo 0.00 37.22 0.00

Renosterveld 3.94 0.00 0.00

Sparse Arid Woodland 0.30 0.00 0.00

Sour Grassland 27.71 0.00 0.00

Succulent Karoo 0.00 27.35 0.00

Sweet Grassland 12.56 0.00 0.00

Thicket 0.00 35.43 0.00

Water bodies 0.00 0.00 0.00

Total number 1346 223 26

No settlements were given the highest overall fire likelihood of ‘Almost certain’ (Forsyth et al.,

2010), but almost half of them fall into the ‘Likely’ and almost a third into the ‘Possible’ class.

This indicates that management measures aimed at reducing the fire risks should be

implemented in more than 80% of the settlements that were assessed. This is not particularly

surprising given that most of the population and the settlements are located in the wetter

eastern and southern parts of the country where the vegetation produces sufficient fuel for

fires at quite short intervals.

34

Table 10: Towns categorised by the likelihood of a wildfire occurring within the Wildland-Urban Interface

Fire danger

(likelihood of occurrence)

Fire return interval

(years)

Number of settlements

(% of total)

Likely 10 781 (48.9%)

Possible 15 528 (33.1%)

Unlikely 25 8 (0.5%)

Rare 100 279 (17.48%)

Total 1 596

Most settlements showed substantial variations in fire hazard (likelihood) within the outer

buffer (Figure 4) suggesting that there is opportunity for vegetation management aimed at fuel

reduction and fire protection. In this example, Clarens is clearly exposed to a much greater

risk of fire than Bethlehem.

Figure 4: Spatial variations in the likelihood of wildfires occurring within the exterior 1 km buffer on WUI for

Bethlehem and Clarens in the eastern Free State.

4.2 Economic and social consequences

The economic consequences of a wildfire occurring within the 1 km interior buffer of the

Wildland-Urban Interface are generally classed as ‘Moderate’ with most of the rest being

‘Minor’ (Table 11). Only a few settlements had a rating of ‘Major’ and just one was

‘Catastrophic’ and these should be the focus of investments and actions aimed at reducing

the consequences.

35

Table 11: Towns categorised by the economic consequences of a wildfire occurring within the outer 1 km buffer of the

WUI. For descriptions of the consequence classes see Table 6

Economic consequence category

Number of settlements Percent of total

Catastrophic 1 0.06%

Major 104 6.52%

Moderate 1 100 68.92%

Minor 344 21.55%

Insignificant 47 2.94%

Total 1 596

The process of deriving the classification used in the interior 1 km wide buffer is best

understood using a settlement as an example. An analysis of the land-cover classes within

the buffer shows that there is a varied mix (Error! Reference source not found.5).

Figure 5: Land-cover classes present in the internal buffer of the Wildland-Urban Interface of a single South Africa

settlement based on the 2013-14 national land-cover (GTI, 2015).

When these are regrouped into the economic consequence classes it is clear that most of the

buffer comprises land-cover classes that fall into the ‘Insignificant’ to ‘Moderate’ economic

36

consequence classes (Error! Reference source not found.6), and that the ‘Major’ class is

resent but mainly confined to a few portions of the buffer. Overall, the ‘Moderate’ economic

consequence class occupies the greatest proportion of the buffer and so the economic

consequence of a wildfire in the interior buffer of the settlement WUI is characterised as

‘Moderate’ (Error! Reference source not found.).

Figure 6: Land-cover classes for a single town categorised by levels of economic consequence present within the

internal buffer of the Wildland-Urban Interface of a single South Africa settlement.

37

Figure 7: Economic risks aggregated to the dominant risk class occurring within the interior buffer of the Wildland-

Urban Interface of a single town in South Africa.

4.3 Fire-danger days under current and future climates

The number of very high fire-danger days per year was calculated for 1 596 towns and for

three time periods: baseline (1961 – 1990), near future (2021 - 2050) and far future (2070 -

2099). The outputs for the current climate or baseline (1961 – 1990) show that there is a high

number of very high fire-danger days in the arid north-western and the western coastal parts

of South Africa (Error! Reference source not found.), decreasing steeply to 0-25 days in

uch of the southern part of the Western Cape, almost all of the Eastern Cape and KwaZulu-

Natal and Mpumalanga and the western Free State. Most of the Northern Cape is Karoo

shrublands which almost never experience fires (Archibald et al., 2010b; Le Maitre et al.,

2014b), but a strip extending from the northern part of this province through Limpopo will

experience 50-100 very high fire-danger days and is arid savanna and grassland. These

38

vegetation types can experience fires after growing seasons with higher than normal rainfall

(Archibald et al., 2010a; Le Maitre et al., 2014b).

Figure 8: Classes of very high fire-danger days for southern Africa under the current climate (1961-1990).

The projections for both the near future and the far future show a southward and eastward

expansion of the occurrence of >25 very high fire-danger days per year and a contraction in

the areas experiencing 0-25 days per year. The most marked shifts are in the Free State,

Western Cape, Eastern Cape, North West and Limpopo provinces.

39

Figure 9: Classes of very high fire-danger days for southern Africa under the climate projects for the near future (2021-

2050).

Figure 10: Classes of very high fire-danger days for southern Africa under the climate projects for the far future (2021-

2050).

40

4.4 Cederberg Local Municipality

The annual total of very high fire-danger days was then assessed for each of the 1 596 towns

in the study using the settlement polygons. The outputs of this assessment can be illustrated

for settlements in the Cederberg Municipality in the Western Cape (Figure 11) for both the

current climate (Figure 12) and for the near (Figure 13) and far future (Figure 14).

Figure 11: The Cederberg Local Municipality showing the location of the important settlements and the variable

topography from the nearly flat West Coast lowlands to the rugged interior, particularly the Cederberg Mountains east

of Clanwilliam.

The inland towns show a progressively worsening situation while the coastal towns remain in

the same class throughout. For example, Clanwilliam currently experiences 81-140 very high

fire-danger days a year, will experience about the same in the near future, but will experience

an increase to 141-183 such days in the far future. This municipality is located in the area of

relatively large numbers of very high fire-danger days in the hinterland of the West Coast in

the Western Cape and towns in the southern part of the province (e.g. George) will experience

much less change.

41

Figure 12: The number of very high fire-danger days for each the towns within the Cederberg Municipality under the

current climate (1961 to 1990

Figure 13: The number of very high fire-danger days predicted for the near future (2021 to 2050) for each of the towns

within the Cederberg Municipality.

42

Figure 14: The number of very high fire-danger days predicted for the far future (2070 to 2099) for each the towns

within the Cederberg Municipality.

4.5 Dihlabeng Local Municipality

Another example is the Dihlabeng Municipality in the eastern Free State which includes the

town of Bethlehem (Error! Reference source not found.5).

Figure 15: Dihlabeng Municipality in the eastern Free State showing the main town of Bethlehem and the other small

settlements. The southern boundary of the municipality is the border with Lesotho.

43

The fire-danger (likelihood of a fire) varies markedly between the different settlements, with

Bethlehem and Rosendal classed as ‘Possible’ and the other settlements as ‘Likely’,(but the

economic consequences (‘Moderate’) are the same for all of them (Figure 17).

Figure 16: The fire hazard (likelihood) of the different settlements in the Dihlabeng Municipality in the eastern Free

State.

Figure 17: The economic consequences of wildfires in the settlements in the Dihlabeng Municipality in the eastern

Free State.

44

Figure 18: The economic risk associated with the occurrence of a wildfire in the settlements in the Dihlabeng

Municipality in the eastern Free State.

Under the current climate the settlements within the Dihlabeng Municipality experience few

very high fire-danger days per year (Error! Reference source not found.19) with Paul Roux

nd Rosendal experiencing slightly more than the other settlements. In the near future Paul

Roux shifts up one class to 11-25 days per year and in the far future all of the settlements shift

up by one to two classes.

45

Figure 19: The number of very high fire-danger days for each of the towns within the Dihlabeng Municipality under the

current climate (1961 to 1990).

Figure 20: The number of very high fire-danger days for each of the towns within the Dihlabeng Municipality under the

climate projected for the near future (2021 to 2050).

Figure 21: The number of very high fire-danger days for each of the towns within the Dihlabeng Municipality under the

climate projected for the far future (2070 to 2099).

46

4.6 Fire occurrence

The analysis of fire occurrences in each settlement’s 1 km wide outside buffer shows some

interesting differences among the fire ecology types (Error! Reference source not found.2).

lightly more than half the settlements have no recorded fires in their buffers. This does not

mean that there have not been any fires because some fires, especially ones less than about

12 ha may not be detected and others, especially in Fynbos and Renosterveld, may not be

detected at all (De Klerk et al., 2012; Le Maitre et al., 2014b).

Those with no fires include all those in the Succulent Karoo, Nama Karoo, Sparse Arid

Woodland and Forest fire ecology types, which is what would be expected given that fires are

expected to be rare (<1 in 100 years) in these fire ecology types. In contrast the settlements

in the grassland and woodland fire ecology types are much more likely to have had fires, with

more than 75% of those in the Sour Grassland, and nearly 60% of those in Moist Woodland,

having had at least one fire. This is also in line with the expectations given that fires can occur

every year in these fire ecology types. Slightly more than half of the settlements in

Renosterveld have had more than one fire which differs from Fynbos where most have not

had a fire.

Table 12: The numbers of fires recorded (i.e. the number of times a MODIS pixel was detected as burnt) in the outside

buffer of each settlement.

Ecotypes Maximum number of fires in pixel in

0 1 2 ≥3

Arid Woodland 164 87 7 7

Coastal Grassland 39 17 3 1

Forest 5 0 0 0

Fynbos 95 52 1 2

Grassy Nama Karoo 19 1 1 0

Moist Woodland 109 107 33 22

Nama Karoo 79 3 1 0

Renosterveld 25 23 5 0

Sour Grassland 85 213 34 41

Sparse Arid Woodland 4 0 0 0

Succulent Karoo 61 0 0 0

Sweet Grassland 61 87 14 7

Thicket 71 6 2 0

Water bodies 1 0 0 0

Total number of towns 818 596 101 80

47

The percentage of the settlement buffer that has burnt over time shows a somewhat different

picture. In the Sour Grassland and Moist Woodland, although the outside buffer may only have

had one or two fires (Error! Reference source not found.2), those fires have cumulatively

urnt quite a large proportion of that buffer and thus posed a greater fire hazard than is indicated

by the occurrence alone. So, 17% of the settlements in the Sour Grassland and 9% of those

in the Moist Woodland have has more than 50% of their buffer area burnt.

Table 13: The percentage of area of the outside buffer burnt over a 15 year time period for each settlement in each fire

ecology type.

Fire ecology type

Percentage class

0 0.01 – 15.00

15.01 – 25.00

25.01 – 50.00

>50.00

Arid Woodland 164 56 16 26 3

Coastal Grassland 39 16 2 1 2

Forest 5 0 0 0 0

Fynbos 93 37 7 10 3

Grassy Nama Karoo 19 2 0 0 0

Moist Woodland 109 70 27 42 24

Nama Karoo 79 3 0 1 0

Renosterveld 25 17 5 5 1

Sour Grassland 187 58 16 26 60

Sparse Arid Woodland 4 0 0 0 0

Succulent Karoo 61 0 0 0 0

Sweet Grassland 62 39 23 28 17

Thicket 71 8 0 0 0

Water bodies 1 0 0 0 0

Total number of towns 818 341 128 200 110

48

5 CONCLUSION

This study extends beyond the 2010 assessment at the local authority level (Forsyth et al.,

2010) because it explicitly assesses the fire hazard and the consequences or vulnerability at

the WUI. The analysis shows that the majority of the settlements are located in fire ecology

types that require fire and burn regularly. This means that the communities living in these

settlements need to be adequately prepared to deal with such fires. This will involve a range

of measures, including fuel reduction to reduce fire intensity and to exclude fires to protect

vulnerable infrastructure. They also need to establish procedures to ensure that their level of

preparedness to respond increases in line with increasing fire-danger ratings.

People living in settlements that do not have Fire Protection Associations should form them

so that they can be better prepared to protect themselves from the negative impacts of

unplanned fires. There is a lot of information available on the internet for people who are

interested in reducing their fire risks, including:

Firewise South Africa: www.firewisesa.org.za

Working on Fire: https://workingonfire.org/

Fynbos Fire: http://fynbosfire.org.za/

Kishugu: http://kishugu.com/firescape-your-home-before-its-too-late/

This is only an initial assessment aimed at providing an overview of the relative fire hazard

and fire risks for many settlements across the country. It needs to be supplemented with a

more detailed local authority and settlement level assessment of the risk incorporating local

knowledge and consultation with the affected communities (Holloway et al., 2010). We

strongly recommend that local authorities which have settlements that are exposed to fire risks

initiate and conduct such assessments in a participatory and consultative fashion.

49

6 REFERENCES

Alexandre, P.M., Stewart, S.I., Keuler, N.S., Clayton, M.K., Mockrin, M.H., Bar-Massada, A.,

Syphard, A.D., Radeloff, V.C., 2016. Factors related to building loss due to wildfires in

the conterminous United States. Ecol. Appl. 26, 2323–2338.

Archibald, S., Lehmann, C.E.R., Belcher, C.M., Bond, W.J., Bradstock, R.A., Daniau, A.-L.,

Dexter, K.G., Forrestel, E.J., Greve, M., He, T., Higgins, S.I., Hoffmann, W.A., Lamont,

B.B., McGlinn, D.J., Moncrieff, G.R., Osborne, C.P., Pausas, J.G., Price, O., Ripley,

B.S., Rogers, B.M., Schwilk, D.W., Simon, M.F., Turetsky, M.R., Van der Werf, G.R.,

Zanne, A.E., 2018. Biological and geophysical feedbacks with fire in the Earth system.

Environ. Res. Lett. 13, 033003.

Archibald, S., Scholes, R.J., 2007. Leaf green-up in a semi-arid African savanna -separating

tree and grass responses to environmental cues. J. Veg. Sci. 18, 583–594.

Archibald, S.A., Nickless, A., Govender, N., Scholes, R.J., Lehsten, V., 2010a. Climate and

the inter-annual variability of fire in southern Africa: a meta-analysis using long-term

field data and satellite-derived burnt area data. Glob. Ecol. Biogeogr. 19, 794–809.

Archibald, S.A., Roy, D.P., van Wilgen, B.W., Scholes, R.J., 2009. What limits fire? An

examination of drivers of burnt area in Southern Africa. Glob. Chang. Biol. 15, 613–

630.

Archibald, S.A., Scholes, R.J.A., Roy, D.P.C., Roberts, G.D., Boschetti, L.E., 2010b.

Southern African fire regimes as revealed by remote sensing. Int. J. Wildl. Fire 19,

861–878.

Baeza, M.J., De Luı́s, M., Raventós, J., Escarré, A., 2002. Factors influencing fire behaviour

in shrublands of different stand ages and the implications for using prescribed burning

to reduce wildfire risk. J. Environ. Manage. 65, 199–208.

Batidzirai, B., Valk, M., Wicke, B., Junginger, M., Daioglou, V., Euler, W., Faaij, A.P.C.,

2016. Current and future technical, economic and environmental feasibility of maize

and wheat residues supply for biomass energy application: Illustrated for South Africa.

Biomass and Bioenergy 92, 106–129.

Blanchi, R., Leonard, J.E., Leicester, R.H., 2006. Lessons learnt from post-bushfire surveys

at the urban interface in Australia. For. Ecol. Manage. 234, S139.

50

Bond, W.J., 1997. Fire. In: Cowling, R.M., Richardson, D.M., Pierce, S.M. (Eds.), Vegetation

of Southern Africa. Cambridge University Press, Cambridge, United Kingdom, pp.

421–446.

Bond, W.J., Honig, M., Maze, K.M., 1999. Seed size and seedling emergence: an allometric

relationship and some ecological implications. Oecologia 120, 132–136.

Bond, W.J., Midgley, G.F., Woodward, F.I., 2003. What controls South African vegetation —

climate or fire ? South African J. Bot. 69, 1–13.

Bond, W.J., Scott, A.C., 2010. Fire and the spread of flowering plants in the Cretaceous.

New Phytol. 188, 1137–1150.

Buthelezi, N.L.P., Mutanga, O., Rouget, M., Sibanda, M., 2016. A spatial and temporal

assessment of fire regimes on different vegetation types using MODIS burnt area

products. Bothalia 46, 9 pages.

Conedera, M., Tonini, M., Oleggini, L., Vega Orozco, C., Leuenberger, M., Pezzatti, G.B.,

2015. Geospatial approach for defining the Wildland-Urban Interface in the Alpine

environment. Comput. Environ. Urban Syst. 52, 10–20.

de Klerk, H.M., Wilson, A.M., Steenkamp, K., 2012. Evaluation of satellite-derived burned

area products for the fynbos, a Mediterranean shrubland. Int. J. Wildl. Fire 21, 36.

Deshmukh, I.K., 1984. A common relationship between precipitation and grassland peak

biomass for East and southern Africa. Afr. J. Ecol. 22, 181–186.

Edgeley, C.M., Paveglio, T.B., 2017. Community recovery and assistance following large

wildfires: The case of the Carlton Complex Fire. Int. J. Disaster Risk Reduct. 1–0.

Ellery, W.N., Scholes, R.J., Scholes, M.C., 1995. The distribution of sweetveld and sourveld

in South Africa’s grassland biome in relation to environmental factors. African J. Range

Forage Sci. 12, 38–45.

Everson, T.M., Smith, F.R., Everson, C.S., 1985. Characteristics of fire behaviour in the

montane grasslands of natal. J. Grassl. Soc. South. Africa 2, 13–21.

Everson, T.M., Van Wilgen, B.W., Everson, C.S., 1988. Adaptation of a model for rating fire

danger in the Natal Drakensberg. S. Afr. J. Sci. 84, 44–49.

Forsyth, G.G., Kruger, F.J., Le Maitre, D.C., 2010. National Veldfire Risk Assessment :

Analysis of Exposure of Social , Economic and Environmental Assets. Report No:

CSIR/NRE/ECO/ER/2010/0023/C, Natural Resources and the Environment, CSIR,

Stellenbosch for the Department of Agriculture, Fisheries and Forestry, Pretoria.

51

Forsyth, G.G., Le Maitre, D.C., 2015. Risk to communities at the Wildland-Urban Interface.

Report No. CSIR/NRE/ECOS/ER/2015/0075/B, Natural Resources and the

Environment, CSIR, Stellenbosch.

Forsyth, G.G., Le Maitre, D.C., van Wilgen, B.W., 2015. GEF Fynbosfire Project Information

Management, Fire Research & Modelling and Fire Risk Assessment Component:

Climate. Progress report on the GEF FynbosFire Project, prepared by Natural

Resources and the Environment, CSIR, Stellenbosch.

Geldenhuys, C.J., 1994. Bergwind fires and the location pattern of forest patches in the

southern Cape landscape, South Africa. J. Biogeogr. 21, 49–62.

Ghermandi, L., Beletzky, N.A., de Torres Curth, M.I., Oddi, F.J., 2016. From leaves to

landscape: A multiscale approach to assess fire hazard in wildland-urban interface

areas. J. Environ. Manage. 183, 925–937.

Gill, A.M., Stephens, S.L., Cary, G.J., 2013. The worldwide ‘“wildfire”’ problem. Ecol. Appl.

23, 438–454.

Gill, A.M., Zylstra, P., 2005. Flammability of Australian forests. Aust. For. 68, 87–93.

Goldammer, J.G., de Ronde, C., 2004. Wildland Fire Management Handbook for Sub-

Sahara Africa. Global Fire Monitoring Center (GFMC), Freiburg, Germany.

Govender, N., Trollope, W.S.W., van Wilgen, B.W., 2006. The effect of fire season, fire

frequency, rainfall and management on fire intensity in savanna vegetation in South

Africa. J. Appl. Ecol. 43, 748–758.

GTI, 2015. 2013 - 2014 South African National Land-Cover Dataset. Data User Report and

MetaData. Version 05 (DEA Open Access), Geoterraimage, Pretoria.

Hély, C., Dowty, P.R., Alleaume, S., Caylor, K.K., Korontzi, S., Swap, R.J., Shugart, H.H.,

Justice, C.O., 2003. Regional fuel load for two climatically contrasting years in

southern Africa. J. Geophys. Res. Atmos. 108, n/a-n/a.

Herrero-Corral, G., Jappiot, M., Bouillon, C., Long-Fournel, M., 2012. Application of a

geographical assessment method for the characterization of wildland–urban interfaces

in the context of wildfire prevention: A case study in western Madrid. Appl. Geogr. 35,

60–70.

Hoare, D.B., Frost, P., 2004. Phenological description of natural vegetation in southern

Africa using remotely-sensed vegetation data. Appl. Veg. Sci. 7, 19–28.

52

Holloway, A., Fortune, G., Vimbai, C., 2010. RADAR Western Cape 2010. Risk and

Development Annual Review. Cape Town.

Jayiya, T.P., Moses, G., Forsyth, G.G., Van Wilgen, B.W., 2004. The effects of invasion by

alien shrubs and trees on the fuel properties of ecosystems in the Western Cape ,

South Africa. Natural Resources and the Environment, CSIR, Stellenbosch.

Keeley, J.E., 2009. Fire intensity, fire severity and burn severity: a brief review and

suggested usage. Int. J. Wildl. Fire 18, 116.

Kotzé, I., Beukes, H., van den Berg, E., Newby, T., 2010. National Invasive Alien Plant

Survey. Report No. GW/A/2010/21, Agricultural Research Council – Institute for Soil,

Climate and Water, Pretoria.

Kraaij, T., Cowling, R.M., Van Wilgen, B.W., 2013. Fire regimes in eastern coastal fynbos:

Imperatives and thresholds in managing for diversity. Koedoe 55, 1–9.

Kraaij, T., Wilgen, B.W. Van, 2014. Drivers, ecology, and management of fire in fynbos. In:

Allsopp, N., Colville, J.F., Verboom, G.A. (Eds.), Fynbos: Ecology, Evolution, and

Conservation of a Megadiverse Region. Oxford University Press, Oxford, UK, pp. 47–

72.

Lampin-Maillet, C., Jappiot, M., Long, M., Bouillon, C., Morge, D., Ferrier, J.-P., 2010.

Mapping wildland-urban interfaces at large scales integrating housing density and

vegetation aggregation for fire prevention in the South of France. J. Environ. Manage.

91, 732–741.

Le Maitre, D.C., 2015. Fynbos Fuel Handbook. Information management, Fire research &

modelling, and fire risk assessment component, GEF FynbosFire Project.

Le Maitre, D.C., Forsyth, G.G., Engelbrecht, F., Landman, W., Frost, P., Van Wilgen, B.,

2015. Climate, fire regimes and fire danger. Information Management, Fire Research &

Modelling and Fire Risk Assessment Component, GEF Fynbosfire Project. Report No

CSIR/NRE/ECOS/ER/2015/0079/B, Natural Resources and the Environment, CSIR,

Stellenbosch.

Le Maitre, D.C., Kotzee, I.M., O’Farrell, P.J., 2014a. Impacts of land-cover change on the

water flow regulation ecosystem service: Invasive alien plants, fire and their policy

implications. Land use policy 36, 171–181.

53

Le Maitre, D.C., Kruger, F.J., Forsyth, G.G., 2014b. Interfacing ecology and policy:

Developing an ecological framework and evidence base to support wildfire

management in South Africa. Austral Ecol. 39, 424–436.

McConnachie, M.M., van Wilgen, B.W., Richardson, D.M., Ferraro, P.J., Forsyth, A.T., 2015.

Estimating the effect of plantations on pine invasions in protected areas: a case study

from South Africa. J. Appl. Ecol. 52, 110–118.

Midgley, G.F., Chapman, R.A., Hewitson, B., De Wit, M., Ziervogel, G., Mukheibir, P., Van

Niekerk, L., Tadross, M., Van Wilgen, B.W., Kgope, B., Morant, P.D., Scholes, R.J.,

Forsyth, G.G., 2005. A Status Quo, Vulnerability and Adaptation Assessment of the

Physical and Socio-Economic Effects of Climate Change in the Western Cape.

Technical report, CSIR Report No. ENV-S-C 2005-073. Cape Town: Department of

Environmental Affairs and Development Planning, Provincial Government of the

Western Cape.

Mucina, L., Rutherford, M.C., 2006. The vegetation of South Africa, Lesotho and Swaziland.

Strelitzia 19. South African National Biodiversity Institute, Pretoria, Pretoria.

Naudé, A., Badenhorst, W., Zietsman, L., Van Huyssteen, E., Maritz, J., 2007. Geospatial

Analysis Platform – Version 2: Technical overview of the mesoframe methodology and

South African Geospatial Analysis Platform. CSIR Report number:

CSIR/BE/PSS/IR/2007/0104/B, Built Environment, CSIR, Pretoria.

Nel, J.L., Le Maitre, D.C., Nel, D.C., Reyers, B., Archibald, S., van Wilgen, B.W., Forsyth,

G.G., Theron, A.K., O’Farrell, P.J., Kahinda, J.-M.M., Engelbrecht, F.A., Kapangaziwiri,

E., van Niekerk, L., Barwell, L., 2014. Natural hazards in a changing world: a case for

ecosystem-based management. PLoS One 9, e95942.

Noble, I.R., Gill, A.M., Bary, G.A. V., 1980. McArthur’s fire-danger meters expressed as

equations. Austral Ecol. 5, 201–203.

Parrini, F., Erasmus, B.F.N., 2009. Spatial and temporal heterogeneity of phenology patterns

in Kruger National Park, South Africa: different drivers for different areas. In:

Proceedings, 33rd International Symposium on Remote Sensing of Environment,

ISRSE 2009 2009, Stresa, Italy. pp. 627–631.

Phillips, J.F. V, 1931. Forest succession and ecology in the Knysna region. Memoirs of the

Botanical Survey of South Africa No. 14, Botanical Research Institute, Pretoria.

Richardson, D.M., Bond, W.J., 1985. Determinants of plant distribution: evidence from pine

invasions. Am. Nat. 138, 639–668.

54

Richardson, D.M., Williams, P.A., Hobbs, R.J., 2000. Pine invasions in the Southern

Hemisphere: Determinants of spread and invadability. J. Biogeogr. 21, 511–527.

Roy, D.P., Boschetti, L., 2009. Southern Africa Validation of the MODIS, L3JRC, and

GlobCarbon Burned-Area Products. IEEE Trans. Geosci. Remote Sens. 47, 1032–

1044.

Schwilk, D.W., 2015. Dimensions of plant flammability. New Phytol. 206, 486–8.

Scott, D.F., 1993. The hydrological effects of fire in South African mountain catchments. J.

Hydrol. 150, 409–432.

Scott, D.F., 1997. The contrasting effects of wildfire and clearfelling on the hydrology of a

small catchment. Hydrol. Process. 11, 543–555.

Scott, D.F., 2000. Soil wettability in forested catchments in South Africa; as measured by

different methods and as affected by vegetation cover and soil characteristics. J.

Hydrol. 231–232, 87–104.

Scott, D.F., Versfeld, D.B., Lesch, W., 1998. Erosion and sediment yield in relation to

afforestation and fire in the mountains of the Western Cape Province, South Africa.

South African Geogr. J. 80, 52–59.

Shakesby, R., Doerr, S., 2006. Wildfire as a hydrological and geomorphological agent.

Earth-Science Rev. 74, 269–307.

Shakesby, R.A., 2011. Post-wildfire soil erosion in the Mediterranean: Review and future

research directions. Earth-Science Rev. 105, 71–100.

Snyman, H.J., Fouche, H.A., 1991. Production and water-use efficiency of semi-arid

grasslands of South Africa as affected by veld condition and rainfall. Water SA 17,

263–268.

Syphard, A.D., Bar Massada, A., Butsic, V., Keeley, J.E., 2013. Land use planning and

wildfire: development policies influence future probability of housing loss. PLoS One 8,

e71708.

Tainton, N.M., 1999. Veld management in southern Africa. Natal University Press,

Pietermaritzburg.

Teie, W.C., 2009. Fire Manager’s Handbook on Veld and Forest Fires : Strategy, Tactics and

Safety, 2nd ed. Southern African Insititute of Forestry, Menlo Park, Pretoria.

55

Tsela, P., Wessels, K., Botai, J., Archibald, S., Swanepoel, D., Steenkamp, K., Frost, P.,

2014. Validation of the Two Standard MODIS Satellite Burned-Area Products and an

Empirically-Derived Merged Product in South Africa. Remote Sens. 6, 1275–1293.

Van Huyssteen, E., Biermann, S., Naude, A., Le Roux, A., 2009. Advances in spatial

analysis to support a more nuanced reading of the South African space economy.

Urban Forum 20, 195–214.

Van Wilgen, B.W., 1984. Adaptation of the United States fire danger rating system to fynbos

conditions. I: A fuel model for fire danger rating in the fynbos biome. South African For.

J. 129, 61–65.

Van Wilgen, B.W., Archibald, S., Kirton, A., Forsyth, G.., Das, S., Khuluse, S., Schmitz, P.,

Frost, P., Steenkamp, K., McFerren, G., 2010. Changes in the extent, frequency and

severity of wildfires. In: Scholes, R.J. (Ed.), Global Change Risk Analysis:

Understanding and Reducing Key Risks to Ecosystem Services Associated with

Climate Change in South Africa. Report number CSIR/NRE/ECO/IR/2010/0047/C,

CSIR, Pretoria, pp. 63–81.

van Wilgen, B.W., Everson, C.S., Trollope, W.S.W., 1990. Fire Management in Southern

Africa: Some Examples of Current Objectives, Practices, and Problems. In:

Goldammer, J.G. (Ed.), Fire in the Tropical Biota. Springer-Verlag, Berlin, pp. 179–

215.

Van Wilgen, B.W., Higgins, K.B., Bellstedt, D.U., 1990. The role of vegetation structure and

fuel chemistry in excluding fire from forest patches in the fire-prone fynbos shrublands

of South Africa. J. Ecol. 78, 210–222.

van Wilgen, B.W., Richardson, D.M., 1985. The Effects of Alien Shrub Invasions on

Vegetation Structure and Fire Behaviour in South African Fynbos Shrublands: A

Simulation Study. J. Appl. Ecol. 22, 955–966.

van Wilgen, B.W., Richardson, D.M., 2012. Three centuries of managing introduced conifers

in South Africa: Benefits, impacts, changing perceptions and conflict resolution. J.

Environ. Manage. 106, 56–68.

van Wilgen, B.W., Richardson, D.M., 2014. Challenges and trade-offs in the management of

invasive alien trees. Biol. Invasions 16, 721–734.

van Wilgen, B.W., Scholes, R.J., 1997. The vegetation and fire regimes of southern

hemisphere Africa. In: van Wilgen, B.W., Andreae, M.O., Goldammer, J.G., Lindesay,

56

J.A. (Eds.), Fire in Southern African Savannas. Ecological and Atmospheric

Perspectives. Witwatersrand University Press, Johannesburg., pp. 27–46.

Van Wilgen, B.W., Scott, D.F., 2001. Managing fires on the Cape Peninsula: dealing with the

inevitable. J. Mediterr. Ecol. 2, 197–208.

Wessels, K., Steenkamp, K., von Maltitz, G., Archibald, S., 2011. Remotely sensed

vegetation phenology for describing and predicting the biomes of South Africa. Appl.

Veg. Sci. 14, 49–66.

Whitecross, M.A., Witkowski, E.T.F., Archibald, S., 2017. Savanna tree-grass interactions: A

phenological investigation of green-up in relation to water availability over three

seasons. South African J. Bot. 108, 29–40.

Whitman, E., Rapaport, E., Sherren, K., 2013. Modeling fire susceptibility to delineate

wildland-urban interface for municipal-scale fire risk management. Environ. Manage.

52, 1427–39.

Willis, C., van Wilgen, B., Tolhurst, K., Everson, C., D’Abreton, P., Pero, L., Fleming, G.,

2001. The Development of a National Fire Danger Rating System for South Africa.

Pretoria